1 /* Input parser for Bison
3 Copyright (C) 1984, 1986, 1989, 1992, 1998, 2000, 2001, 2002, 2003,
4 2005, 2006, 2007, 2009 Free Software Foundation, Inc.
6 This file is part of Bison, the GNU Compiler Compiler.
8 This program is free software: you can redistribute it and/or modify
9 it under the terms of the GNU General Public License as published by
10 the Free Software Foundation, either version 3 of the License, or
11 (at your option) any later version.
13 This program is distributed in the hope that it will be useful,
14 but WITHOUT ANY WARRANTY; without even the implied warranty of
15 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 GNU General Public License for more details.
18 You should have received a copy of the GNU General Public License
19 along with this program. If not, see <http://www.gnu.org/licenses/>. */
28 #include "conflicts.h"
32 #include "muscle-tab.h"
36 #include "scan-gram.h"
37 #include "scan-code.h"
39 static void check_and_convert_grammar (void);
41 static symbol_list
*grammar
= NULL
;
42 static bool start_flag
= false;
43 merger_list
*merge_functions
;
45 /* Was %union seen? */
46 bool union_seen
= false;
49 bool tag_seen
= false;
51 /* Should rules have a default precedence? */
52 bool default_prec
= true;
54 /*-----------------------.
55 | Set the start symbol. |
56 `-----------------------*/
59 grammar_start_symbol_set (symbol
*sym
, location loc
)
62 complain_at (loc
, _("multiple %s declarations"), "%start");
67 startsymbol_location
= loc
;
73 /*------------------------------------------------------------------------.
74 | Return the merger index for a merging function named NAME. Records the |
75 | function, if new, in MERGER_LIST. |
76 `------------------------------------------------------------------------*/
79 get_merge_function (uniqstr name
)
88 head
.next
= merge_functions
;
89 for (syms
= &head
, n
= 1; syms
->next
; syms
= syms
->next
, n
+= 1)
90 if (UNIQSTR_EQ (name
, syms
->next
->name
))
92 if (syms
->next
== NULL
)
94 syms
->next
= xmalloc (sizeof syms
->next
[0]);
95 syms
->next
->name
= uniqstr_new (name
);
96 /* After all symbol type declarations have been parsed, packgram invokes
97 record_merge_function_type to set the type. */
98 syms
->next
->type
= NULL
;
99 syms
->next
->next
= NULL
;
100 merge_functions
= head
.next
;
105 /*-------------------------------------------------------------------------.
106 | For the existing merging function with index MERGER, record the result |
107 | type as TYPE as required by the lhs of the rule whose %merge declaration |
108 | is at DECLARATION_LOC. |
109 `-------------------------------------------------------------------------*/
112 record_merge_function_type (int merger
, uniqstr type
, location declaration_loc
)
115 merger_list
*merge_function
;
121 type
= uniqstr_new ("");
124 for (merge_function
= merge_functions
;
125 merge_function
!= NULL
&& merger_find
!= merger
;
126 merge_function
= merge_function
->next
)
128 aver (merge_function
!= NULL
&& merger_find
== merger
);
129 if (merge_function
->type
!= NULL
&& !UNIQSTR_EQ (merge_function
->type
, type
))
131 complain_at (declaration_loc
,
132 _("result type clash on merge function `%s': <%s> != <%s>"),
133 merge_function
->name
, type
, merge_function
->type
);
134 complain_at (merge_function
->type_declaration_location
,
135 _("previous declaration"));
137 merge_function
->type
= uniqstr_new (type
);
138 merge_function
->type_declaration_location
= declaration_loc
;
141 /*--------------------------------------.
142 | Free all merge-function definitions. |
143 `--------------------------------------*/
146 free_merger_functions (void)
148 merger_list
*L0
= merge_functions
;
151 merger_list
*L1
= L0
->next
;
158 /*-------------------------------------------------------------------.
159 | Parse the input grammar into a one symbol_list structure. Each |
160 | rule is represented by a sequence of symbols: the left hand side |
161 | followed by the contents of the right hand side, followed by a |
162 | null pointer instead of a symbol to terminate the rule. The next |
163 | symbol is the lhs of the following rule. |
165 | All actions are copied out, labelled by the rule number they apply |
167 `-------------------------------------------------------------------*/
169 /* The (currently) last symbol of GRAMMAR. */
170 static symbol_list
*grammar_end
= NULL
;
172 /* Append SYM to the grammar. */
174 grammar_symbol_append (symbol
*sym
, location loc
)
176 symbol_list
*p
= symbol_list_sym_new (sym
, loc
);
179 grammar_end
->next
= p
;
185 /* A null SYM stands for an end of rule; it is not an actual
194 assign_named_ref (symbol_list
*p
, named_ref
*named_ref
)
196 symbol
*sym
= p
->content
.sym
;
198 if (named_ref
->id
== sym
->tag
)
200 warn_at (named_ref
->loc
,
201 _("duplicated symbol name for %s ignored"),
203 named_ref_free (named_ref
);
206 p
->named_ref
= named_ref
;
210 /* The rule currently being defined, and the previous rule.
211 CURRENT_RULE points to the first LHS of the current rule, while
212 PREVIOUS_RULE_END points to the *end* of the previous rule (NULL). */
213 static symbol_list
*current_rule
= NULL
;
214 static symbol_list
*previous_rule_end
= NULL
;
217 /*----------------------------------------------.
218 | Create a new rule for LHS in to the GRAMMAR. |
219 `----------------------------------------------*/
222 grammar_current_rule_begin (symbol
*lhs
, location loc
,
223 named_ref
*lhs_named_ref
)
227 /* Start a new rule and record its lhs. */
229 previous_rule_end
= grammar_end
;
231 p
= grammar_symbol_append (lhs
, loc
);
233 assign_named_ref(p
, lhs_named_ref
);
235 current_rule
= grammar_end
;
237 /* Mark the rule's lhs as a nonterminal if not already so. */
238 if (lhs
->class == unknown_sym
)
240 lhs
->class = nterm_sym
;
244 else if (lhs
->class == token_sym
)
245 complain_at (loc
, _("rule given for %s, which is a token"), lhs
->tag
);
249 /*----------------------------------------------------------------------.
250 | A symbol should be used if either: |
251 | 1. It has a destructor. |
252 | 2. --warnings=midrule-values and the symbol is a mid-rule symbol |
253 | (i.e., the generated LHS replacing a mid-rule action) that was |
254 | assigned to or used, as in "exp: { $$ = 1; } { $$ = $1; }". |
255 `----------------------------------------------------------------------*/
258 symbol_should_be_used (symbol_list
const *s
)
260 if (symbol_destructor_get (s
->content
.sym
)->code
)
262 if (warnings_flag
& warnings_midrule_values
)
263 return ((s
->midrule
&& s
->midrule
->action_props
.is_value_used
)
264 || (s
->midrule_parent_rule
265 && symbol_list_n_get (s
->midrule_parent_rule
,
266 s
->midrule_parent_rhs_index
)
267 ->action_props
.is_value_used
));
271 /*----------------------------------------------------------------.
272 | Check that the rule R is properly defined. For instance, there |
273 | should be no type clash on the default action. |
274 `----------------------------------------------------------------*/
277 grammar_rule_check (const symbol_list
*r
)
281 If there is an action, then there is nothing we can do: the user
282 is allowed to shoot herself in the foot.
284 Don't worry about the default action if $$ is untyped, since $$'s
285 value can't be used. */
286 if (!r
->action_props
.code
&& r
->content
.sym
->type_name
)
288 symbol
*first_rhs
= r
->next
->content
.sym
;
289 /* If $$ is being set in default way, report if any type mismatch. */
292 char const *lhs_type
= r
->content
.sym
->type_name
;
293 const char *rhs_type
=
294 first_rhs
->type_name
? first_rhs
->type_name
: "";
295 if (!UNIQSTR_EQ (lhs_type
, rhs_type
))
296 warn_at (r
->location
,
297 _("type clash on default action: <%s> != <%s>"),
300 /* Warn if there is no default for $$ but we need one. */
302 warn_at (r
->location
,
303 _("empty rule for typed nonterminal, and no action"));
306 /* Check that symbol values that should be used are in fact used. */
308 symbol_list
const *l
= r
;
310 for (; l
&& l
->content
.sym
; l
= l
->next
, ++n
)
311 if (! (l
->action_props
.is_value_used
312 || !symbol_should_be_used (l
)
313 /* The default action, $$ = $1, `uses' both. */
314 || (!r
->action_props
.code
&& (n
== 0 || n
== 1))))
317 warn_at (r
->location
, _("unused value: $%d"), n
);
319 warn_at (r
->location
, _("unset value: $$"));
325 /*-------------------------------------.
326 | End the currently being grown rule. |
327 `-------------------------------------*/
330 grammar_current_rule_end (location loc
)
332 /* Put an empty link in the list to mark the end of this rule */
333 grammar_symbol_append (NULL
, grammar_end
->location
);
334 current_rule
->location
= loc
;
338 /*-------------------------------------------------------------------.
339 | The previous action turns out the be a mid-rule action. Attach it |
340 | to the current rule, i.e., create a dummy symbol, attach it this |
341 | mid-rule action, and append this dummy nonterminal to the current |
343 `-------------------------------------------------------------------*/
346 grammar_midrule_action (void)
348 /* Since the action was written out with this rule's number, we must
349 give the new rule this number by inserting the new rule before
352 /* Make a DUMMY nonterminal, whose location is that of the midrule
353 action. Create the MIDRULE. */
354 location dummy_location
= current_rule
->action_props
.location
;
355 symbol
*dummy
= dummy_symbol_get (dummy_location
);
356 symbol_list
*midrule
= symbol_list_sym_new (dummy
, dummy_location
);
358 /* Remember named_ref of previous action. */
359 named_ref
*named_ref
= current_rule
->action_props
.named_ref
;
361 /* Make a new rule, whose body is empty, before the current one, so
362 that the action just read can belong to it. */
365 /* Attach its location and actions to that of the DUMMY. */
366 midrule
->location
= dummy_location
;
367 code_props_rule_action_init (&midrule
->action_props
,
368 current_rule
->action_props
.code
,
369 current_rule
->action_props
.location
,
371 code_props_none_init (¤t_rule
->action_props
);
373 if (previous_rule_end
)
374 previous_rule_end
->next
= midrule
;
378 /* End the dummy's rule. */
379 midrule
->next
= symbol_list_sym_new (NULL
, dummy_location
);
380 midrule
->next
->next
= current_rule
;
382 previous_rule_end
= midrule
->next
;
384 /* Insert the dummy nonterminal replacing the midrule action into
385 the current rule. Bind it to its dedicated rule. */
386 grammar_current_rule_symbol_append (dummy
, dummy_location
, named_ref
);
387 grammar_end
->midrule
= midrule
;
388 midrule
->midrule_parent_rule
= current_rule
;
389 midrule
->midrule_parent_rhs_index
= symbol_list_length (current_rule
->next
);
392 /* Set the precedence symbol of the current rule to PRECSYM. */
395 grammar_current_rule_prec_set (symbol
*precsym
, location loc
)
397 symbol_class_set (precsym
, token_sym
, loc
, false);
398 if (current_rule
->ruleprec
)
399 complain_at (loc
, _("only one %s allowed per rule"), "%prec");
400 current_rule
->ruleprec
= precsym
;
403 /* Attach dynamic precedence DPREC to the current rule. */
406 grammar_current_rule_dprec_set (int dprec
, location loc
)
409 warn_at (loc
, _("%s affects only GLR parsers"), "%dprec");
411 complain_at (loc
, _("%s must be followed by positive number"), "%dprec");
412 else if (current_rule
->dprec
!= 0)
413 complain_at (loc
, _("only one %s allowed per rule"), "%dprec");
414 current_rule
->dprec
= dprec
;
417 /* Attach a merge function NAME with argument type TYPE to current
421 grammar_current_rule_merge_set (uniqstr name
, location loc
)
424 warn_at (loc
, _("%s affects only GLR parsers"), "%merge");
425 if (current_rule
->merger
!= 0)
426 complain_at (loc
, _("only one %s allowed per rule"), "%merge");
427 current_rule
->merger
= get_merge_function (name
);
428 current_rule
->merger_declaration_location
= loc
;
431 /* Attach SYM to the current rule. If needed, move the previous
432 action as a mid-rule action. */
435 grammar_current_rule_symbol_append (symbol
*sym
, location loc
,
436 named_ref
*named_ref
)
439 if (current_rule
->action_props
.code
)
440 grammar_midrule_action ();
441 p
= grammar_symbol_append (sym
, loc
);
443 assign_named_ref(p
, named_ref
);
446 /* Attach an ACTION to the current rule. */
449 grammar_current_rule_action_append (const char *action
, location loc
,
450 named_ref
*named_ref
)
452 if (current_rule
->action_props
.code
)
453 grammar_midrule_action ();
454 /* After all symbol declarations have been parsed, packgram invokes
455 code_props_translate_code. */
456 code_props_rule_action_init (¤t_rule
->action_props
, action
, loc
,
457 current_rule
, named_ref
);
461 /*---------------------------------------------------------------.
462 | Convert the rules into the representation using RRHS, RLHS and |
464 `---------------------------------------------------------------*/
469 unsigned int itemno
= 0;
470 rule_number ruleno
= 0;
471 symbol_list
*p
= grammar
;
473 ritem
= xnmalloc (nritems
+ 1, sizeof *ritem
);
475 /* This sentinel is used by build_relations in gram.c. */
478 rules
= xnmalloc (nrules
, sizeof *rules
);
483 symbol
*ruleprec
= p
->ruleprec
;
484 record_merge_function_type (p
->merger
, p
->content
.sym
->type_name
,
485 p
->merger_declaration_location
);
486 rules
[ruleno
].user_number
= ruleno
;
487 rules
[ruleno
].number
= ruleno
;
488 rules
[ruleno
].lhs
= p
->content
.sym
;
489 rules
[ruleno
].rhs
= ritem
+ itemno
;
490 rules
[ruleno
].prec
= NULL
;
491 rules
[ruleno
].dprec
= p
->dprec
;
492 rules
[ruleno
].merger
= p
->merger
;
493 rules
[ruleno
].precsym
= NULL
;
494 rules
[ruleno
].location
= p
->location
;
495 rules
[ruleno
].useful
= true;
496 rules
[ruleno
].action
= p
->action_props
.code
;
497 rules
[ruleno
].action_location
= p
->action_props
.location
;
499 /* If the midrule's $$ is set or its $n is used, remove the `$' from the
500 symbol name so that it's a user-defined symbol so that the default
501 %destructor and %printer apply. */
502 if (p
->midrule_parent_rule
503 && (p
->action_props
.is_value_used
504 || symbol_list_n_get (p
->midrule_parent_rule
,
505 p
->midrule_parent_rhs_index
)
506 ->action_props
.is_value_used
))
507 p
->content
.sym
->tag
+= 1;
509 /* Don't check the generated rule 0. It has no action, so some rhs
510 symbols may appear unused, but the parsing algorithm ensures that
511 %destructor's are invoked appropriately. */
513 grammar_rule_check (p
);
515 for (p
= p
->next
; p
&& p
->content
.sym
; p
= p
->next
)
519 /* Don't allow rule_length == INT_MAX, since that might
520 cause confusion with strtol if INT_MAX == LONG_MAX. */
521 if (rule_length
== INT_MAX
)
522 fatal_at (rules
[ruleno
].location
, _("rule is too long"));
524 /* item_number = symbol_number.
525 But the former needs to contain more: negative rule numbers. */
527 symbol_number_as_item_number (p
->content
.sym
->number
);
528 /* A rule gets by default the precedence and associativity
529 of its last token. */
530 if (p
->content
.sym
->class == token_sym
&& default_prec
)
531 rules
[ruleno
].prec
= p
->content
.sym
;
534 /* If this rule has a %prec,
535 the specified symbol's precedence replaces the default. */
538 rules
[ruleno
].precsym
= ruleprec
;
539 rules
[ruleno
].prec
= ruleprec
;
541 /* An item ends by the rule number (negated). */
542 ritem
[itemno
++] = rule_number_as_item_number (ruleno
);
543 aver (itemno
< ITEM_NUMBER_MAX
);
545 aver (ruleno
< RULE_NUMBER_MAX
);
551 aver (itemno
== nritems
);
553 if (trace_flag
& trace_sets
)
554 ritem_print (stderr
);
557 /*------------------------------------------------------------------.
558 | Read in the grammar specification and record it in the format |
559 | described in gram.h. All actions are copied into ACTION_OBSTACK, |
560 | in each case forming the body of a C function (YYACTION) which |
561 | contains a switch statement to decide which action to execute. |
562 `------------------------------------------------------------------*/
567 /* Initialize the symbol table. */
570 /* Construct the accept symbol. */
571 accept
= symbol_get ("$accept", empty_location
);
572 accept
->class = nterm_sym
;
573 accept
->number
= nvars
++;
575 /* Construct the error token */
576 errtoken
= symbol_get ("error", empty_location
);
577 errtoken
->class = token_sym
;
578 errtoken
->number
= ntokens
++;
580 /* Construct a token that represents all undefined literal tokens.
581 It is always token number 2. */
582 undeftoken
= symbol_get ("$undefined", empty_location
);
583 undeftoken
->class = token_sym
;
584 undeftoken
->number
= ntokens
++;
586 gram_in
= xfopen (grammar_file
, "r");
588 gram__flex_debug
= trace_flag
& trace_scan
;
589 gram_debug
= trace_flag
& trace_parse
;
590 gram_scanner_initialize ();
593 /* Set front-end %define variable defaults. */
594 muscle_percent_define_default ("lr.keep-unreachable-states", "false");
597 /* IELR would be a better default, but LALR is historically the
599 muscle_percent_define_default ("lr.type", "LALR");
600 lr_type
= muscle_percent_define_get ("lr.type");
601 if (0 != strcmp (lr_type
, "canonical LR"))
602 muscle_percent_define_default ("lr.default-reductions", "all");
604 muscle_percent_define_default ("lr.default-reductions", "accepting");
608 /* Check front-end %define variables. */
610 static char const * const values
[] = {
611 "lr.type", "LALR", "IELR", "canonical LR", NULL
,
612 "lr.default-reductions", "all", "consistent", "accepting", NULL
,
615 muscle_percent_define_check_values (values
);
618 if (! complaint_issued
)
619 check_and_convert_grammar ();
625 /*-------------------------------------------------------------.
626 | Check the grammar that has just been read, and convert it to |
628 `-------------------------------------------------------------*/
631 check_and_convert_grammar (void)
633 /* Grammar has been read. Do some checking. */
635 fatal (_("no rules in the input grammar"));
637 /* Report any undefined symbols and consider them nonterminals. */
638 symbols_check_defined ();
640 /* If the user did not define her ENDTOKEN, do it now. */
643 endtoken
= symbol_get ("$end", empty_location
);
644 endtoken
->class = token_sym
;
645 endtoken
->number
= 0;
646 /* Value specified by POSIX. */
647 endtoken
->user_token_number
= 0;
650 /* Find the start symbol if no %start. */
655 node
!= NULL
&& symbol_is_dummy (node
->content
.sym
);
658 for (node
= node
->next
;
659 node
!= NULL
&& node
->content
.sym
!= NULL
;
664 grammar_start_symbol_set (node
->content
.sym
,
665 node
->content
.sym
->location
);
668 /* Insert the initial rule, whose line is that of the first rule
669 (not that of the start symbol):
671 accept: %start EOF. */
673 symbol_list
*p
= symbol_list_sym_new (accept
, empty_location
);
674 p
->location
= grammar
->location
;
675 p
->next
= symbol_list_sym_new (startsymbol
, empty_location
);
676 p
->next
->next
= symbol_list_sym_new (endtoken
, empty_location
);
677 p
->next
->next
->next
= symbol_list_sym_new (NULL
, empty_location
);
678 p
->next
->next
->next
->next
= grammar
;
684 aver (nsyms
<= SYMBOL_NUMBER_MAXIMUM
&& nsyms
== ntokens
+ nvars
);
686 /* Assign the symbols their symbol numbers. Write #defines for the
687 token symbols into FDEFINES if requested. */
690 /* Scan rule actions after invoking symbol_check_alias_consistency (in
691 symbols_pack above) so that token types are set correctly before the rule
692 action type checking.
694 Before invoking grammar_rule_check (in packgram below) on any rule, make
695 sure all actions have already been scanned in order to set `used' flags.
696 Otherwise, checking that a midrule's $$ should be set will not always work
697 properly because the check must forward-reference the midrule's parent
698 rule. For the same reason, all the `used' flags must be set before
699 checking whether to remove `$' from any midrule symbol name (also in
703 for (sym
= grammar
; sym
; sym
= sym
->next
)
704 code_props_translate_code (&sym
->action_props
);
707 /* Convert the grammar into the format described in gram.h. */
710 /* The grammar as a symbol_list is no longer needed. */
711 symbol_list_free (grammar
);